I appreciate the simplicity of Dust's setup. It's accessible to anyone while being very advanced and efficient. Many tasks can be accomplished with Dust, making it very simple and maintainable. The different agents created and used are very traceable, which is great for team collaboration. I also like the constant addition of features. Review collected by and hosted on G2.com.

At certain times, I noticed that there was no possibility for video creation in Dust, which leads me to use other tools. Moreover, Dust does not allow interaction with a browser, a feature that is important to me and that I use with other tools. These are the two main points that come to mind. Review collected by and hosted on G2.com.

I really like how Dust lets me set things up with pods and with the different models or agents I can define myself. That flexibility gives me more control and better organization than other chatbots or applications I’ve tried. The assistance feature is especially helpful for creating agents with clear, specific instructions, and for sharing workflows between them when I need to reuse the same approach.

Pods are also excellent for keeping my projects organized and collecting all the documents for each project in one place. On top of that, I like the search tools—especially semantic and hierarchical search—which make it easier for me to find what I need and manage information effectively. Overall, Dust feels intuitive and solid, and for my use cases it’s been straightforward to use. Review collected by and hosted on G2.com.

I think there are a few organizational areas that could be improved, especially the features for finding things as projects get bigger. For me, organization and having some control over how everything is arranged are the most important parts here. Review collected by and hosted on G2.com.

I appreciate being able to create assistants and automate processes with Dust. We configure assistants for specific use cases, which allows us to save time and be more productive, as the assistants capitalize on our best practices and knowledge. This frees up time for other tasks and for skill development. The initial installation of Dust is easy, quick, and fun. Review collected by and hosted on G2.com.

I am having difficulties with the ease of use of Dust. There is also an issue with the locking of prompts: when I create an agent and share it with my teams, they have the ability to modify and view the prompts. This can lead to leaks and changes that we won't be able to track, which means a waste of time. I also find that the tracking and orchestration of agents could be improved, as well as the calculation of consumption. Review collected by and hosted on G2.com.

What I like the most:

Having access to all the LLMs. It's really a considerable advantage. I really like switching between different models according to my needs.

The management of agents is really very interesting, everything is in one place, it's easy to manage.

The UX/UI of the DUST site is very simple, very easy to use, there's nothing to complain about. Review collected by and hosted on G2.com.

Regarding the agents, I would really like to be able to connect my agents to each other so that they can call/launch themselves. This is one of the major points missing in Dust.

The integration of certain MCPs can sometimes be a bit lengthy.

Some triggers, notably Jira, still do not work.

I would really like Dust to be able to intervene on my machine / on certain websites when I authorize it... Review collected by and hosted on G2.com.

What I appreciate about DUST is the topic of governance, GDPR compliance, and the security of my data to prevent training LLMs.

The ability to collaborate with colleagues on the same subject through LLMs is a big plus.

This has allowed my colleagues to go beyond their scope of action before using Dust. Review collected by and hosted on G2.com.

I would like to be able to create agents and sell them to clients. Unfortunately, an agent necessarily has access to the company's global knowledge, which does not allow me to invite a client to use an agent that I would have created for them. Review collected by and hosted on G2.com.

The simplicity of handling. It's ultra intuitive, although it's better for optimal adoption to appoint a "responsible person" in the company who can create the agents for everyone (and especially the connections with the different tools). This is actually a point that follows: the connection in 3 clicks via the browser is pure bliss. Review collected by and hosted on G2.com.

We are going to switch to using tokens/credits and it quickly becomes intimidating. I also have quite a bit of trouble talking to someone when I have questions. Review collected by and hosted on G2.com.

I really appreciate how Dust helps me streamline and automate various parts of my job as a Technical Support Engineer, especially in managing tasks like creating, escalating, and prioritizing incidents and service requests. I love being able to swipe many of these tasks to Dust without trade-offs, finding it customizable to perform at higher volumes with more speed and similar accuracy. The clean and user-friendly UX/UI is a big plus, and the product is rich with features, including the prompt addition of the latest models. I enjoy Dust's granularity with different models being more performant in areas like speed and tool calling. The tool's integration capabilities with main connectors like Slack, Drive, and support for Custom MCPs like Zoho Desk and Snowflake are very beneficial. Features like Pods and sidekick, as well as a chrome extension that unlocks browser use, really expand AI workflow possibilities. The initial setup was super easy too, requiring little to no effort on my part. Review collected by and hosted on G2.com.

I find there are a few options missing like better filters for conversations and a few options for bulk updates. Also, having options for computer use and to run scripts with agents would be nice. Review collected by and hosted on G2.com.

I love the ability of Dust agents to connect to the same tools (CRM, Drive, etc.) with shared (or personal) connections and to share the same context for the entire team. This allows everyone to share the same foundations, KPI definitions, etc. I also appreciate that Dust is very easy to start with, which makes it simple to implement because you can begin with something very basic and then add complexity over time, especially by creating more complex agents and connecting them to more systems. Review collected by and hosted on G2.com.

I would say that there aren't a lot of things to improve, but I think the ability to have better logs on what happened with the agents would be useful. Rather than just being like an AI chat, it would be better to have agents more in the background, running on their own, and showing when there is a problem rather than being considered as normal chats. This would transform Dust from a contextual chat platform to a true AI agent orchestrator. Review collected by and hosted on G2.com.

I greatly appreciate the ease of use of Dust. It is truly a major asset, especially with features like Sidekick that help even the less experienced to create their own agents and choose the right models and data sources. This allows each team to become autonomous, even for those who are less familiar with AI best practices. Moreover, the Dust community is a real advantage, thanks to an engaged community and ambassadors who share best practices, which helps us all improve our usage. Review collected by and hosted on G2.com.

What could be improved is to help us control our credit consumption (ban LLMs, set limits on agents, ...) but this is being developed by the Dust teams, so it should be coming soon. Review collected by and hosted on G2.com.

I like Dust's collaborative approach to building, as it allows for quickly reaching high levels of adoption. Being able to use multiple providers brings a certain agility to the creation process. I appreciate Dust's very favorable time-to-value, which facilitates the start of the AI dynamic. Moreover, new features are frequent, which is a great asset. Review collected by and hosted on G2.com.

I find it difficult to imagine a scalable AI project with Dust, even though the computer skill has allowed for some improvement in possibilities. There is really a lack of an A/B testing interface, which makes the testing phase laborious since we have to create duplicate agents and test independently. Overall, the building interface could be revised, particularly by highlighting 'advanced' parameters like model selection. The ease of setup is excellent, but if not well managed, it can quickly lead to dysfunctional and/or extremely under-optimized and expensive agents. The change in pricing is too abrupt, which does not make the platform scalable for many users, 90% of whom are not ready for the transition and are considering leaving Dust. Review collected by and hosted on G2.com.